Deep Learning: Mapshift-Prediction based on (depre- cated) map priors for online map perception (deutsch/english)

Research topic/area

Deep Learning in Autonomous Driving

Type of thesis

Bachelor / Master

Start time

01.05.2025

Application deadline

30.09.2025

Duration of the thesis

4 - 8 Monate

Description

Current state-of-the-art map construction methods such as MapTRv2 use sensor data (360° surround view camera setup and LiDAR) to construct high definition (HD) maps. These methods extract features from the sensor data and transform them into a Bird’s Eye View (BEV) representation and derive maps in polyline representation using transformer-based architectures. However, the quality of the predicted HD map depends on precise sensor data with correct calibration and an accurate vehicle localization. In real-world scenarios there is always some noise in the localization and sensor data so that a systematic correction is beneficial.

The goal of this thesis is to investigate the possibilities of a learned mapshift predictor. Therefore, the map construction model MapTRv2 should be extended by a second prediction head, which estimates the shift with respect to the x- and y-axis and a possible rotation ϕ of the predicted map compared to the ground truth map. For a reliable mapshift estimate also prior knowledge is needed. For that reason, using different types of prior knowledge such as information from deprecated maps or other information should be evaluated. The dataset that will be used in this thesis is Argoverse 2

Requirement

Requirements for students

• Knowledge in Python, PyTorch and Deep Learning
• Knowledge in Linux and Maps is a plus
• Independent working style and interest in learning new things

Faculty departments

• Engineering sciences
  Electrical engineering & information technologies
  Informatics
  Mechanical engineering
  Mechatronics & information technologies
• Natural sciences and Technology
  Mathematics
  Physics

Supervision

Title, first name, last name

Jonas, Merkert

Organizational unit

Institut für Mess- und Regelungstechnik

Email address

jonas.merkert@kit.edu

Link to personal homepage/personal page

Website

Application via email

Application documents

• Curriculum vitae
• Grade transcript
• Certificate of enrollment

E-Mail Address for application
Senden Sie die oben genannten Bewerbungsunterlagen bitte per Mail an jonas.merkert@kit.edu


Back